I am fitting a Cox regression model using the coxph() function of the survival package in R. I have 15 predictors. One of the predictors is treatment (given / not given). The treatment is a well-established predictor, but I am interested in studying whether or not this is related to the other predictors (i.e., does treatment benefit survival only for certain categories of the other predictors, but not others)? If I attempt to add 14 interactions terms with my 15 predictors, some of the coefficients are infinite. How else can I approach my question?
2
-
$\begingroup$ What is your sample size and proportion of censoring? This will need a lot of events to fit. $\endgroup$– juodCommented Aug 11, 2019 at 22:35
-
$\begingroup$ 400 patients and 2:1 (censored to uncensored). The predictors are informed by the literature, but some predicators are more important than others (if I have to be selective). $\endgroup$– AbeontCommented Aug 11, 2019 at 22:45
Add a comment
|